Oncogenic roles of the SETDB2 histone methyltransferase in gastric cancer

SETDB2 is a histone H3 lysine 9 (H3K9) tri-methyltransferase that is involved in transcriptional gene silencing. Since it is still unknown whether SETDB2 is linked to carcinogenesis, we studied alterations and functions of SETDB2 in human gastric cancers (GCs). SETDB2 protein was highly expressed in 30 of 72 (41.7%) primary GC tissues compared with their normal counterparts by immunohistochemistry. SETDB2 overexpression was significantly associated with the late stage of GCs (P<0.05) and poor prognosis of GC patients (P<0.05). The GC cell lines with SETDB2 knockdown and overexpression significantly decreased and increased cell proliferation, migration and invasion, respectively (P<0.05). Knockdown of SETDB2 in MKN74 and MKN45 cells reduced global H3K9 tri-methylation (me3) levels. Microarray analysis indicated that expression of WWOX and CADM1, tumor suppressor genes, was significantly enhanced in MKN74 cells after SETDB2 knockdown. Chromatin immunoprecipitation assays showed that the H3K9me3 levels at the promoter regions of these two genes corresponded to the SETDB2 expression levels in GC cells. Moreover, ectopic SETDB2 protein was recruited to their promoter regions. Our data suggest that SETDB2 is associated with transcriptional repression of WWOX and CADM1, and hence overexpression of SETDB2 may contribute to GC progression.


INTRODUCTION
Gastric cancer (GC) is the third leading cause of death from cancer in the world, and the incidence is still high [1]. Genetic and epigenetic alterations are involved in gastric carcinogenesis; for example, mutations of the CDH1 and p53 tumor suppressor genes (TSGs), and hypermethylation at the CpG island promoter region of TSGs have been detected in GCs [2,3].
Tri-methylation mark of histone H3 lysine 4 (H3K4me3) is involved in transcriptional activation of genes, whereas tri-methylation marks of H3K9 (H3K9me3) and K27 (H3K27me3) are associated with gene silencing in the transcriptional process. These histone methylations are catalyzed by histone methyltranseferases (HMTs) containing an evolutionally served SET (SuVar3-9, enhancer of Zeste, Trithorax) domain [4]. Among them, SETDB1, SUV39H1 and EZH2 are reported to be frequently over-expressed in various cancers, such as GCs, those of which functions are shown as oncogenic activities [5][6][7][8]. We have recently reported that expression of SET7/9, a histone H3K4 monomethyltransferase, was reduced in advanced GCs [9]. Thus, expression changes of histone modification genes are known to play an important role in tumor development.

Research Paper
SETDB2 have been characterized as H3K9 trimethyltrasferases that contain a bifurcated SET domain, a pre-SET domain and a Methyl-CpG binding domain [10,11]. Setdb2 restricts dorsal organizer territory and regulates structural left-right asymmetry of visceral organs and nervous system by suppressing the activity of fibroblast growth factor 8 in zebrafish embryo [10]. Moreover, it was reported that knockdown of Setdb2 in zebrafish embryo induced abnormal convergence and extension movements during gastrulation, resulting in anterior-posterior shortening [12], indicating an essential role in embryonic development. Although SETDB2 is known to be associated with chromosomal condensation and segregation through H3K9 tri-methylation (me3) regulation during mitosis [11], the molecular mechanism by which SETDB2 mediates transcriptional gene silencing remains unclear. Furthermore, there has not been reported the evidence presented of SETDB2 alterations in cancers. In this study, we aimed to clarify whether SETDB2 is linked to carcinogenesis. Since we found that SETDB2 was often overexpressed in both primary GC tissues and GC cell lines, we further analyzed its biological functions in GC cells.

Overexpression of SETDB2 in primary GC tissues
We analyzed the expression of SETDB2 protein in a set of 72 primary GC tissues by immunohistochemical staining (IHC). SETDB2 expression was not detected in any non-cancerous stomach tissues, whereas SETDB2 was strongly expressed in the nuclei of 30 GC tissues (41.7%, Figure 1A). Western blot analysis showed that SETDB2 protein was expressed in four of 12 primary GC tissues (33.3 %) but not in any of the non-cancerous stomach tissues tested ( Figure 1B). To evaluate the significance of SETDB2 alteration in primary GCs, we studied the relationship between SETDB2 overexpression and clinicopathological factors of GCs. SETDB2 was significantly overexpressed in advanced GC tissues compared with early GC tissues (P<0.05, Table 1). A significant difference was also found between SETDB2 expression and pTNM stages (Stage I vs. II-IV, P<0.05). However, there were no significant correlations between SETDB2 expression and age, gender, histological classification or lymph node metastasis. Furthermore, SETDB2 expression was significantly associated with worse survival by Kaplan-Meier analysis (P<0.05, Figure  1C). Thus, SETDB2 overexpression might be potentially associated with clinical progression of GCs.

Analyses of SETDB2 expression in GC cell lines
We next examined SETDB2 protein expression levels in 13 GC cell lines using Western blot analysis.
By densitometric quantification of the SETDB2 bands, we subsequently classified these cell lines into SETDB2 high- (HSC57, MKN7, MKN45 and MKN74) and low-(AGS, GCIY, TGBC11TKB, HSC39, HSC44PE, HSC59 NUGC3, NUGC4 and KATOIII) expression groups, which were determined by a cut-off value of 0.50 (Figure 2A). To evaluate the biological significance of SETDB2 expression, the effect of knockdown of SETDB2 expressing using RNAinterference was analyzed. SETDB2 protein expression was efficiently suppressed by transfection of SETDB2 siRNAs (siSETDB2) into MKN74 and MKN45 cells compared with its expression in negative control siRNA-transfected cells ( Figure 2B). These two cell lines in which SETDB2 was knocked down showed significant inhibition of cell proliferation compared with the cells transfected with the negative control siRNA (P<0.05, Figure 2C). Moreover, knockdown of SETDB2 markedly decreased the cell migration and invasion rates of MKN45 cells (P<0.05, Figure  2D and 2E). In addition, we analyzed the effect of SETDB2 overexpression in AGS and NUGC3 cells by transfection with the SETDB2/pcDNA3 expression vector ( Figure 2F). Cell growth, migration and invasion were enhanced in the GC cells with SETDB2 overexpression compared with the GC cells transfected with empty vector (P<0.05, Figure 2G-2I). Scratch assay also demonstrated that cell migration was significantly faster in AGS and NUGC3 cells with SETDB2 overexpression than those with empty vector transfection (P<0.05, Supplementary Figure S1A).

The relationship between SETDB2 expression and global H3K9 methylation in GC cells
To investigate whether or not altered SETDB2 expression evokes H3K9 methylation change in GC cells, we compared global H3K9 methylation levels in MKN74 and MKN45 cells between SETDB2 knockdown and control cells. Global H3K9me3 at the protein levels were decreased in both of these cell lines with knockdown of SETDB2 by Western blotting, although global di-(H3K9me2, Figure 3A) and mono-methylation (H3K9me1, data not shown) levels were not changed. To further clarify whether decrease global H3K9me3 levels in our study were independently effects by SETDB2 knockdown or the secondary ones by other H3K9 trimethyltransferase, we examined expression of SETDB1 and SUV39H1 in MKN74 and MKN45 cells with SETDB2 knockdown. The mRNA expression of SETDB1 and SUV39H1 was not changed in MKN74 and MKN45 cells after SETDB2 knockdown (Supplementary Figure  S2), suggesting that SETDB2 itself is correlated with global H3K9me3 levels in these two GC cell lines.

Analyses of SETDB2 target genes in GC cells
To determine SETDB2 target genes, microarray analysis of MKN74 cells with SETDB2 knockdown was www.impactjournals.com/oncotarget   performed. There were 520 genes that showed a greater than 1.5-fold change in expression level in MKN74 cells after SETDB2 knockdown, of which 185 genes were upregulated and 335 genes were down-regulated. Because SETDB2 knockdown led to significant inhibition of cell growth, migration and invasion, we further selected six tumor-or differentiation-related genes (WWOX, CADM1/ TSLC1, CCDC80, NDRG1, CA9 and FZD9) that have been reported to show altered expression in several cancers including GCs [13][14][15][16][17][18]. Expression of WWOX, CADM1/ TSLC1 and CCDC80 genes was up-regulated, while that of NDRG1, CA9 and FZD9 genes was down-regulated in MKN74 cells transfected with SETDB2 siRNA compared with control cells by real-time and conventional RT-PCR analyses ( Figure 3B and Supplementary Figure S3A). These results were consistent with the microarray data. SETDB2 knockdown in MKN45 cells resulted in similar gene expression patterns except for expression of FZD9 ( Figure 3B and Supplementary Figure S3A).

SETDB2 down-regulated the target genes through H3K9me3 in GCs
SETDB2 had an H3K9me3 activity in our study ( Figure 3A), and H3K9me3 is known to be associated with gene silencing [4]. We therefore focused on the three genes (WWOX, CADM1 and CCDC80) whose expression was elevated by SETDB2 knockdown, and we examined how SETDB2 expression inhibited the expression of these genes in GC cells. ChIP assays showed that H3K9me3 levels at the promoter regions of WWOX and CADM1 were greatly decreased in MKN74 and MKN45 cells with SETDB2 siRNA transfection compared with negative control (Figures 4A-4D). In contrast, the H3K9me3 levels at a region approximately +3 kb downstream of the transcriptional start site (TSS) were not changed by SETDB2 siRNA transfection compared with negative control transfection ( Figures 4B and 4D, right panels). Conversely, AGS cells transfected with the SETDB2 expression vector showed increased H3K9me3 levels at the promoter region compared with cells transfected with the empty vector ( Figure 4E, left panels). Furthermore, ChIP assays exhibited that SETDB2 was recruited to the promoter region of WWOX and CADM1 in AGS cells after overexpression of SETDB2 ( Figure 4E, right panels), suggesting that these two genes may be direct targets of SETDB2 in GC cells. Similar interaction of CCDC80 with SETDB2, however, was not detected (data not shown).
We next compared the expression levels of endogenous SETDB2 and its three target genes (WWOX, CADM1 and CCDC80) in 13 GC cell lines using realtime and conventional RT-PCR analyses. The expression levels of WWOX in GC cell lines with high SETDB2 expression were lower than those with low SETDB2 expression, thereby displaying a tendency of WWOX expression to inversely correlate with SETDB2 expression (P=0.058), ( Figure 5A and Supplementary Figure S3B). However, the expression levels of CADM1 and CCDC80 were not correlated with SETDB2 expression levels as assessed by either conventional RT-PCR ( Figure 5A and Supplementary Figure S3B).

Functional analysis of WWOX in GC cells
Since WWOX is known to play a role as a TSG in various cancers [19,20], we studied the relationship between SETDB2 and WWOX. SETDB2 knockdown up-regulated WWOX protein expression in MKN74 and MKN45 cells, and, conversely, SETDB2 overexpression in AGS and NUGC3 cells repressed the protein expression of WWOX ( Figure 5B). To assess whether or not the effects of SETDB2 overexpression in GC cells were mediated by reduced expression of WWOX, we performed knockdown of WWOX by transfection of its siRNA into AGS and NUGC3 GC cells ( Figure 6A). The abilities of cell proliferation, migration and invasion were promoted after knockdown of WWOX in AGS and NUGC3 cells (P<0.05, Figure 6B-6D, Supplementary Figure S1B). It is noted that these effects were similar to those of SETDB2 overexpression in these cells ( Figure 2G-2I, Supplementary Figure S1A).

DISCUSSION
Previous studies have reported the expression changes of the HMT genes in various cancers [21][22][23][24]. The H3K9 tri-methyltransferase genes, SETDB1 and SUV39H1, have been shown to be overexpressed in cancers, such as lung and liver cancers [25][26][27]. Overexpression of these two genes is related to cell proliferation and invasiveness, and Setdb1 accelerated melanoma formation in a transgenic zebrafish model, indicating that they have oncogenic properties [28]. In this study, we showed that the SETDB2 protein was highly expressed in GCs (41.7%), particularly in advanced GCs. Moreover, high SETDB2 expression in GCs was correlated with worse prognosis by Kaplan-Meier analysis (P<0.05). The in vitro studies using SETDB2 knockdown and overexpressing cells of the present study demonstrated that SETDB2 promoted cell proliferation, migration and invasion of GC cells. Thus, SETDB2 expression and function detected in our study were similar to those of SETDB1 and SUV39H1 in cancers [5,6], suggesting that SETDB2 may have oncogenic functions and overexpression of SETDB2 play roles in gastric cancer progression.
Although multiple transcriptional/translational regulatory mechanisms in SETDB1 and SUV39H1 have been identified [26,27], the upstream mechanisms of SETDB2 are unknown. It has been reported that SETDB1 promoter region contains multiple SP1-binding sites and SETDB1 expression was enhanced by SP1, suggesting a significant relation between SETDB1 and TF bindings www.impactjournals.com/oncotarget    [26]. We observed that three putative myc-binding sites were but only one SP1 located at the SETDB2 promoter region by using UCSC Genome Bioinformatics Site (http:// genome.ucsc.edu/cgi-bin/hgGateway). Knockdown of c-myc in MKN74 and MKN45 cells decreased SETDB2 at the mRNA and protein expression levels (Supplementary Figure S4A-S4C), indicating that c-myc may act as a transcriptional regulator of SETDB2 in these two GC cell lines. Besides, amplification of the SETDB1 gene was found in lung and liver cancers [5,25,26], and SETDB1 and SUV39H1 were negatively regulated by miR-29 and miR-125b, respectively [26,27]. Further studies are necessary to clarify the transcriptional and translational regulatory mechanisms of SETDB2 overexpression in GC cells.
The H3K9 methyltransferase genes mediated global and promoter/enhancer region of H3K9 tri-di-and mono-methylation [29][30][31][32]. Knockdown of SETDB2 in MKN74 and MKN45 cells reduced global H3K9me3 ( Figure 3A) but not H3K9me2 and H3K9me1 at the protein level, indicating that SETDB2 participates as an H3K9 tri-methyltransferase [10,11]. High levels of global H3K9me3 protein have been shown in primary cancers including GCs [33,34]. Therefore, we speculate that elevated H3K9me3 levels in GC tissues reported may be mediated in part by SETDB2 overexpression.
In addition, we observed that knockdown of SETDB2 resulted in a decrease in H3K9me3 levels at the promoter regions of WWOX and CADM1 genes, both of whose expression were increased by SETDB2 knockdown. Ectopic SETDB2 in GC cells was recruited to the promoters of these two genes and increased H3K9me3 levels at the regions. Therefore, SETDB2 may repress target gene expression not only through increasing global H3K9me3 levels but also directly binding and inducing H3K9me3 activity on the promoter region of WWOX and CADM1 in GC cells. In contrast, it is possible that CCDC80 may be indirectly regulated by SETDB2, because SETDB2 did not bind to its promoter region.
Regarding the three genes (WWOX, CADM1 and CCDC80) detected as SETDB2 target genes in our study, their loss or reduced expression has been reported in various cancers [35][36][37][38]. Here, loss or low expression of these three genes was found in GC cell lines examined (Figure 5A), and the expression levels of WWOX in GC cell lines showed a tendency to inversely correlate with SETDB2 expression. It is known that WWOX and CADM1 but not CCDC80 have dense CpG islands in their promoter regions and hypermethylation at the promoter regions of WWOX and CADM1 has been reported in cancer cells [39,40]. We examined the methylation status of WWOX and CADM1 genes in 13 GC cell lines. No methylation of WWOX was found in any of the GC cell lines, while severe or partial methylation at the CADM1 promoter was detected in six GC cell lines (data not shown). MKN74 and MKN45 cells exhibited no or partial CADM1 methylation. Our data further support that WWOX and CADM1 may be repressed by SETDB2 in GC cells.
In conclusion, our data indicated that SETDB2 is often over-expressed in GC tissues and cell lines. SETDB2 overexpression significantly accelerated cell proliferation, migration and invasion of GC cells. Moreover, SETDB2 restricted the expression of TSGs such as WWOX, through H3K9me3 at the their promoter regions. Since SETDB2 showed oncogenic functions in our study, it is possible that knockdown of SETDB2 might be an effective therapeutic strategy for GCs.
Cancer tissue samples were collected from 72 GC patients in an affiliated hospital of Tokyo Medical and Dental university. Informed consent was obtained from all of the patients, and the study was approved by the appropriate institutional review committee.

RNA isolation, RT-PCR and real-time RT-PCR analysis
Total RNA was isolated using the Trizol reagent (Thermo Fisher Scientific Inc., Waltham, MA) or the RNeasy plus Minikit (QIAGEN, Valencia, CA). The isolated RNA (1 μg) was reverce-transcribed by using Super Script III Reverse transcriptase (Thermo Fisher Scientific Inc.). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was amplified as an internal control. The primer sequences used are shown in Supplementary Table S1 (Supplementary Material).

Western blot analysis
Total protein was extracted using RIPA buffer and was then electrophoresed on SDS-PAGE. Western blotting was conducted with mouse monoclonal anti-SETDB2 (1:500; M17, Abnova, Taipei, Taiwan), and anti-WWOX

Immunohistochemical staining
Immunohistochemistry (IHC) of formalin-fixed paraffin-embedded tissues was performed as described previously [41]. SETDB2 protein expression was evaluated in a set of 72 primary GC tissues by using rabbit polyclonal anti-SETDB2 (1:100; ab5517, abcam, Cambridge, UK). Sections in which > 10% of the cancer cells showed positive reactivity with SETDB2 were scored as positive. Similary, score of the SETDB2 positivity (>10% of cells) was used in non-cancerous stomach tissues.

Transfection
MKN74 and MKN45 cells with high SETDB2 expression were transfected with SETDB2 siRNAs (final concentration, 50nmol; Hs_SETDB2_0144, Sigma-Aldrich Japan, Ishikari, Japan) or negative control siRNA (Mission_SIC-001, Sigma-Aldrich Japan) using electroporation (Neon TM transfection System, Thermo Fisher Scientific Inc.). AGS and NUGC3 cells with low SETDB2 expression were transfected with the SETDB2/pcDNA3 expression vector (1 μg) or with the empty pcDNA3 vector using the FuGENE HD Transfection Reagent (Promega Corporation, Madison, WI). Transfection of WWOX siRNA (Hs_WWOX_9365, Sigma-Aldrich Japan) was conducted in GC cells by electroporation. Transfected cells were harvested after 48 or 72 hours of culture, and total RNA and protein was then extracted.

Cell growth assay
The effects of SETDB2 knockdown and overexpression on cell proliferation were quantified using the Cell Counting Kit-8 (Dojindo, Kumamoto, Japan). MKN74 (4×10 3 cells/well) and MKN45 (3×10 3 cells/well) cells were transfected with SETDB2 siRNA or negative control siRNA. AGS (1×10 3 cells/well), and NUGC3 (1×10 3 cells/well) cells were transfected with the SETDB2/pcDNA3 or empty pcDNA3 expression vector. Moreover, the cell proliferation rates of AGS and NUGC3 cells transfected with WWOX siRNA were examined. GC cells transfected with siRNA or empty vector were seeded onto 96-well plates, and their cell proliferation was then determined by WST-8 activity.

Transwell migration and invasion assays
MKN45, AGS and NUGC3 cells were transfected with SETDB2 siRNA, WWOX siRNA, full length SETDB2/pcDNA3 or the empty/pcDNA3expression vector. After 48 hours, MKN45 (4×10 3 cells/0.5ml), AGS (3×10 3 cells/0.5ml) or NUGC3 (5×10 3 cells/0.5ml) cells were seeded in a control insert or a Matrigel invasion chamber (BD Biosciences, Franklin Lakes, NJ) [9]. Medium containing 10% FBS in the lower chamber served as chemoattractant. After 48 hours, the non-invaded cells were gently removed with a cotton swab and the migrated and invaded cells located on the lower side of the chamber were stained with crystal violet.

Scratch wound healing assay
At 48 hours after transfection of AGS and NUGC3 cells with the SETDB2 expression vector or with WWOX siRNA, the cells were seeded in 6 well plates at a density of 4×10 5 cells/well and cultured to confluence. A scratch was made through the cell monolayer using a pipet tip. To monitor the migration of cells into the wounded area, photographs of the wounded area were taken immediately after the scratch was made (0 hour) and after then 18 hours later. The relative wound healing range was determined [9].

Microarray analysis
MKN74 cells were transfected with SETDB2 siRNA or negative control siRNA, and were then grown for 48 hours. Total RNA was extracted using RNEasy and cDNA microarray analysis was conducted by DNA Chip Research Inc. (Kanagawa, Japan) using the Agilent Human GE 4x44K v2 Microarray (Platform ID: GPL13497). The raw data have been deposited in NCBI Gene Expression Omnibus (GEO) under the accession number GSE76135.